[MLton-user] Portability and extensions

[Matthew Fluet]

On Tue, Mar 1, 2011 at 5:51 PM, Ivan Tomac <ivan.tomac at gmail.com> wrote:
> In the interest of portability, it might be a good idea to standardize and
> unify some of the common extensions that are present in different SML
> compilers. MLton and SML/NJ in particular are already largely compatible and
> complement each other well - SML/NJ for rapid development and the REPL and
> MLton for compiling the final release of a program.
> Would there be any interest in unifying common extensions under a Platform
> or System structure that exposes a generic interface to things such as weak
> pointers and the garbage collector?

I think that there is always interest.  It seems that every
non-trivial SML project that I've seen includes some compatibility
layer if they intend to build with multiple SML compilers.  For the
most part, when such interfaces are simply renamings or trivial
extensions of existing functionality, they are very easy to add.  The
difficult part is coming up with such an interface that provides a
sufficiently large portion of the extensions that it is useful while
not making it so large that it excludes implementations that don't
have all of the requisite features.

Ivan, if you have what you think is a decent proposal, then feel free
to voice it here and maybe also on the SML/NJ and Poly/ML mailing
lists.